Real-time system and method of monitoring transient operations in continuous casting process for breakout prevention

1. A method for online monitoring of transient operations in a continuous caster and predicting an impending transient-cast breakout or other process abnormality, comprising the following steps: retrieving historical process data of a plurality of selected process variables during a pre-defined transient operation duration, the resulting historical process data covering most of normal transient operation regions and being used to construct a modeling dataset; dividing modeling data in each transient operation into two stages, and synchronizing the modeling data separately with respect to a set of synchronization scales pre-defined by casting speed and casting strand length, respectively, in said two stages to define a series of observations forming a synchronized modeling dataset; performing a multi-way principal component analysis (MPCA) on said synchronized modeling dataset to develop a multivariate statistical model to benchmark normal transient operations; and calculating a loading matrix P, and values of principal components for each transient operation in the modeling dataset; computing test statistics at each observation over a predefined transient operation duration, for each transient operation in the modeling dataset, based on the multivariate statistical model; determining control limits for said test statistics and their contributions from each selected process variable; acquiring on-line process data of each selected process variable from a pre-defined start point to a current time t, in a new transient operation to be monitored; synchronizing the acquired online data based on the synchronization scales defined in the two stages, and predicting future process trajectories for the rest of said transient operation duration, namely, from said current time t to a predefined end point for said transient operation duration to create complete process trajectories; computing test statistics based on the multivariate statistical model for the resulting complete process trajectories of said new transient operation; comparing the test statistics computed from said new transient operation to their corresponding control limits; and generating at least one indication signal indicating whether said new transient operation is statistically different from its historical benchmark of normal operation in a continuous casting process.

2. A method according to claim 1 in which an alarm signal is generated when test statistics exceed their control limits for more than a pre-defined number of consecutive sampling intervals, said alarm signal being indicative of an impending transient-cast breakout or other process abnormality.

3. A method according to claim 2 in which a list of process variables that are most likely contributors to the alarm signal is generated.

4. A method according to claim 1 , in which a transient operation is characterized by slowing down the casting speed, the casting speed of the continuous caster remaining unchanged for a short period of time and finally the casting speed is ramping up gradually back to its normal operating conditions over several minutes.

5. A method according to claim 4 in which the transient operations include SEN changes, flying tundish changes, and product grade changes of a continuous caster.

6. The method according to claim 1 , in which the process variables of a continuous caster are selected from the group comprising: mold thermocouple readings, temperature differences between pre-defined thermocouple pairs, stopper rod position, tundish car net weight, mold cooling water flows, temperature difference between inlet and outlet mold cooling water, and heat flux transferred through each mold face.

7. The method according to claim 6 , wherein the heat flux transferred through each mold face of a continuous caster is calculated from the temperature difference between the inlet and the outlet cooling water flows for each mold face.

8. A method according to claim 1 , in which the duration of a transient operation in a continuous casting process is defined by two stages, namely; a D-stage beginning at the said start point when the casting speed is decreased for a transient operation and ending at an end point when the casting speed is increased from a pre-defined holding speed; and a U-stage beginning at the end of D-stage, and ending at an end point where the length of steel strand cast in U-stage reaches a predetermined length.

9. The method according to claim 8 , in which said casting speed decreases continuously in the D-stage to a desired value which is controlled by an automatic controller.

10. A method according to claim 8 , in which the length of steel strand cast in U-stage is equal to 2.4 meters.

11. A method according to claim 8 , in which said length of steel strand is calculated as the integral of the casting speed over time, and it increases monotonically in the U-stage.

12. A method according to claim 8 , in which process trajectories in D-stage are synchronized based on a set of uniform synchronization scales defined by using the casting speed, and missing data exist at the beginning and/or the end of D-stage due to the applied process trajectory synchronization method.

13. A method according to claim 8 , in which process trajectories in U-stage are synchronized based on a set of non-uniform synchronization scales defined by using the casting strand length such that an online monitoring calculation is conducted more frequently at the beginning of the U-stage than at the end of the U-stage.

14. A method according to claim 1 , in which said multivariate statistical model is developed using MPCA technology and the number of principal components is determined such that a predetermined percentage of operation-to-operation variance existing in the historical modeling dataset has been captured.

15. A method according to claim 1 in which the test statistics are selected from the group consisting of Squared Prediction Error (SPE) and “Hotelling T” (HT).

16. A method according to claim 1 , in which said control limits for said test statistics and contributions of each selected process variable are determined based on the historical data in the modeling dataset, and an adjustable multiplier is used online to adjust said control limits.

17. A method according to claim 16 , in which said multiplier for the control limits of said test statistics has different values in the D- and U-stage.

18. A method according to claim 3 , in which the process variables that are most likely related to the predicted breakouts or process abnormalities are identified by high contribution values, in comparison with their corresponding control limits, at the most current observation.

19. A real-time system for online monitoring of transient operations in a continuous caster and predicting an impending transient-cast breakout or other process abnormality, comprising: a plurality of measurement sensors for obtaining real-time process data of a continuous caster; a data access module to acquire said real-time process data from said sensors, and supply them to other modules in the system, as required; a process state determination module to determine a process state selected from the following: start-up state, shut-down state, run-time state, transient operation state and to select a model calculation module to monitor operations of a continuous caster; a model calculation module, selected by the process state determination module according to said determined process state, to receive said real-time process data, to perform MPCA calculations and to compute test statistics; and a human machine interface for displaying current transient operating conditions according to the determined process state.

20. A system according to claim 19 , in which said transient operation state includes submerged entry nozzle (SEN) changes, flying tundish changes, and product grade changes.

21. A system according to claim 19 , in which said human machine interface provides operating information including any one of the following: identification of said process state, slab tracking identifiers, real-time sensor readings and monitoring information including test statistics, control limits associated with the test statistics and identified process variables that are most likely contributors to generate alarms.

22. A system according to claim 21 , in which said test statistics and contribution values are scaled to [0,1] with respect to corresponding control limits over a defined transient operation duration.

23. A real-time system for online monitoring of transient operations in a continuous caster and predicting an impending transient-cast breakout or other process abnormality comprising: a model development module to receive and divide modeling data from transient operations into two stages, and synchronize the modeling data separately with respect to a set of synchronization scales pre-defined by casting speed and casting strand length, respectively, in said two stages, said model development module defining a series of observations which form a synchronized modeling dataset; to perform a multi-way principal component analysis (MPCA) on said synchronized modeling dataset to develop a multivariate statistical model, which captures operation—operation variance existing in historical data to benchmark normal transient operations; to calculate a loading matrix P, and values of principal components for each transient operation in the modeling dataset; to compute test statistics at each observation over a pre-defined transient operation duration, for each transient operation in the modeling dataset, based on the multivariate statistical model; and to determine control limits for said test statistics and their contributions from selected process variables; a plurality of measurement sensors for acquiring on-line process data of selected process variables from a pre-defined start point to a current time t in a new transient operation being monitored; a synchronization module to synchronize the acquired online process data based on the synchronization scales defined in the two stages, and predict future process trajectories for the rest of said transient operation duration, namely from said current time t to a pre-defined end point for said transient operation duration to create complete process trajectories; a calculation module to compute test statistics based on the multivariate statistical model for the resulting complete process trajectories of said new transient operation and to compare the test statistics computed from said new transient operation to their corresponding control limits; and a human machine interface for said displaying current transient operation conditions.

24. A system according to claim 23 having an alarm for generating an alarm signal when test statistics exceed their control limits for more than a predetermined number of consecutive sampling intervals, said alarm signal being indicative of an impending transient-cast breakout or other process abnormality.

25. A system according to claim 24 in which the human machine interface presents the alarm signal.

26. A system according to claim 24 in which the human machine interface displays a list of process variables that are most likely contributors to the alarm.

27. A system according to claim 23 in which the model development module divides the modeling data in each transient operation into two stages, namely: a D-stage beginning at the said start point when the casting speed is decreased for a transient operation and ending at an end point when the casting speed is increased from a pre-defined holding speed; and a U-stage beginning at the end of D-stage, and ending at an end point where the length of steel strand cast in U-stage reaches a predetermined length.

28. A system according to claim 23 having an adjustable multiplier to adjust control limits online.

29. A system according to claim 23 having means for handling missing data generated by sensor failures, non-functional thermocouples due to changes of continuous caster mold width, and by said synchronization module.